The PCI endpoint device such as Xilinx Alveo PCI card maps the register
spaces from multiple hardware peripherals to its PCI BAR. Normally,
the PCI core discovers devices and BARs using the PCI enumeration process.
There is no infrastructure to discover the hardware peripherals that are
present in a PCI device, and which can be accessed through the PCI BARs.
For Alveo PCI card, the card firmware provides a flattened device tree to
describe the hardware peripherals on its BARs. The Alveo card driver can
load this flattened device tree and leverage device tree framework to
generate platform devices for the hardware peripherals eventually.
Apparently, the device tree framework requires a device tree node for the
PCI device. Thus, it can generate the device tree nodes for hardware
peripherals underneath. Because PCI is self discoverable bus, there might
not be a device tree node created for PCI devices. This patch is to add
support to generate device tree node for PCI devices.
Added a kernel option. When the option is turned on, the kernel will
generate device tree nodes for PCI bridges unconditionally.
Initially, the basic properties are added for the dynamically generated
device tree nodes.
Signed-off-by: Lizhi Hou <[email protected]>
Signed-off-by: Sonal Santan <[email protected]>
Signed-off-by: Max Zhen <[email protected]>
Reviewed-by: Brian Xu <[email protected]>
---
drivers/pci/Kconfig | 12 ++
drivers/pci/Makefile | 1 +
drivers/pci/bus.c | 2 +
drivers/pci/msi/irqdomain.c | 6 +-
drivers/pci/of.c | 71 ++++++++++
drivers/pci/of_property.c | 256 ++++++++++++++++++++++++++++++++++++
drivers/pci/pci-driver.c | 3 +-
drivers/pci/pci.h | 19 +++
drivers/pci/remove.c | 1 +
9 files changed, 368 insertions(+), 3 deletions(-)
create mode 100644 drivers/pci/of_property.c
diff --git a/drivers/pci/Kconfig b/drivers/pci/Kconfig
index 55c028af4bd9..126c31b79718 100644
--- a/drivers/pci/Kconfig
+++ b/drivers/pci/Kconfig
@@ -198,6 +198,18 @@ config PCI_HYPERV
The PCI device frontend driver allows the kernel to import arbitrary
PCI devices from a PCI backend to support PCI driver domains.
+config PCI_DYNAMIC_OF_NODES
+ bool "Device tree node for PCI devices"
+ depends on OF
+ select OF_DYNAMIC
+ help
+ This option enables support for generating device tree nodes for some
+ PCI devices. Thus, the driver of this kind can load and overlay
+ flattened device tree for its downstream devices.
+
+ Once this option is selected, the device tree nodes will be generated
+ for all PCI bridges.
+
choice
prompt "PCI Express hierarchy optimization setting"
default PCIE_BUS_DEFAULT
diff --git a/drivers/pci/Makefile b/drivers/pci/Makefile
index 2680e4c92f0a..cc8b4e01e29d 100644
--- a/drivers/pci/Makefile
+++ b/drivers/pci/Makefile
@@ -32,6 +32,7 @@ obj-$(CONFIG_PCI_P2PDMA) += p2pdma.o
obj-$(CONFIG_XEN_PCIDEV_FRONTEND) += xen-pcifront.o
obj-$(CONFIG_VGA_ARB) += vgaarb.o
obj-$(CONFIG_PCI_DOE) += doe.o
+obj-$(CONFIG_PCI_DYNAMIC_OF_NODES) += of_property.o
# Endpoint library must be initialized before its users
obj-$(CONFIG_PCI_ENDPOINT) += endpoint/
diff --git a/drivers/pci/bus.c b/drivers/pci/bus.c
index 3cef835b375f..8507cc32b61d 100644
--- a/drivers/pci/bus.c
+++ b/drivers/pci/bus.c
@@ -316,6 +316,8 @@ void pci_bus_add_device(struct pci_dev *dev)
*/
pcibios_bus_add_device(dev);
pci_fixup_device(pci_fixup_final, dev);
+ if (pci_is_bridge(dev))
+ of_pci_make_dev_node(dev);
pci_create_sysfs_dev_files(dev);
pci_proc_attach_device(dev);
pci_bridge_d3_update(dev);
diff --git a/drivers/pci/msi/irqdomain.c b/drivers/pci/msi/irqdomain.c
index e9cf318e6670..eeaf44169bfd 100644
--- a/drivers/pci/msi/irqdomain.c
+++ b/drivers/pci/msi/irqdomain.c
@@ -230,8 +230,10 @@ u32 pci_msi_domain_get_msi_rid(struct irq_domain *domain, struct pci_dev *pdev)
pci_for_each_dma_alias(pdev, get_msi_id_cb, &rid);
of_node = irq_domain_get_of_node(domain);
- rid = of_node ? of_msi_map_id(&pdev->dev, of_node, rid) :
- iort_msi_map_id(&pdev->dev, rid);
+ if (of_node && !of_node_check_flag(of_node, OF_DYNAMIC))
+ rid = of_msi_map_id(&pdev->dev, of_node, rid);
+ else
+ rid = iort_msi_map_id(&pdev->dev, rid);
return rid;
}
diff --git a/drivers/pci/of.c b/drivers/pci/of.c
index 196834ed44fe..fb60b04f0b93 100644
--- a/drivers/pci/of.c
+++ b/drivers/pci/of.c
@@ -469,6 +469,8 @@ static int of_irq_parse_pci(const struct pci_dev *pdev, struct of_phandle_args *
} else {
/* We found a P2P bridge, check if it has a node */
ppnode = pci_device_to_OF_node(ppdev);
+ if (of_node_check_flag(ppnode, OF_DYNAMIC))
+ ppnode = NULL;
}
/*
@@ -599,6 +601,75 @@ int devm_of_pci_bridge_init(struct device *dev, struct pci_host_bridge *bridge)
return pci_parse_request_of_pci_ranges(dev, bridge);
}
+#if IS_ENABLED(CONFIG_PCI_DYNAMIC_OF_NODES)
+
+void of_pci_remove_node(struct pci_dev *pdev)
+{
+ struct device_node *dt_node;
+
+ dt_node = pci_device_to_OF_node(pdev);
+ if (!dt_node || !of_node_check_flag(dt_node, OF_DYNAMIC))
+ return;
+ pdev->dev.of_node = NULL;
+
+ of_destroy_node(dt_node);
+}
+
+void of_pci_make_dev_node(struct pci_dev *pdev)
+{
+ struct device_node *parent, *dt_node = NULL;
+ const char *pci_type = "dev";
+ struct of_changeset *cset;
+ const char *full_name;
+ int ret;
+
+ /*
+ * If there is already a device tree node linked to this device,
+ * return immediately.
+ */
+ if (pci_device_to_OF_node(pdev))
+ return;
+
+ /* Check if there is device tree node for parent device */
+ if (!pdev->bus->self)
+ parent = pdev->bus->dev.of_node;
+ else
+ parent = pdev->bus->self->dev.of_node;
+ if (!parent)
+ return;
+
+ if (pci_is_bridge(pdev))
+ pci_type = "pci";
+
+ full_name = kasprintf(GFP_KERNEL, "%pOF/%s@%x,%x", parent, pci_type,
+ PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
+ if (!full_name)
+ goto failed;
+
+ dt_node = of_create_node(parent, full_name, &cset);
+ if (!dt_node)
+ goto failed;
+ kfree(full_name);
+
+ ret = of_pci_add_properties(pdev, cset, dt_node);
+ if (ret)
+ goto failed;
+
+ ret = of_changeset_apply(cset);
+ if (ret)
+ goto failed;
+
+ pdev->dev.of_node = dt_node;
+
+ return;
+
+failed:
+ if (dt_node)
+ of_destroy_node(dt_node);
+ kfree(full_name);
+}
+#endif
+
#endif /* CONFIG_PCI */
/**
diff --git a/drivers/pci/of_property.c b/drivers/pci/of_property.c
new file mode 100644
index 000000000000..cc66fa7517e0
--- /dev/null
+++ b/drivers/pci/of_property.c
@@ -0,0 +1,256 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Copyright (C) 2022, Advanced Micro Devices, Inc.
+ */
+
+#include <linux/pci.h>
+#include <linux/of.h>
+#include <linux/bitfield.h>
+#include <linux/bits.h>
+#include "pci.h"
+
+struct of_pci_addr_pair {
+ __be32 phys_hi;
+ __be32 phys_mid;
+ __be32 phys_lo;
+ __be32 size_hi;
+ __be32 size_lo;
+};
+
+struct of_pci_range {
+ __be32 child_addr_hi;
+ __be32 child_addr_mid;
+ __be32 child_addr_lo;
+ __be32 parent_addr_hi;
+ __be32 parent_addr_mid;
+ __be32 parent_addr_lo;
+ __be32 size_hi;
+ __be32 size_lo;
+};
+
+#define OF_PCI_ADDR_SPACE_CONFIG 0x0
+#define OF_PCI_ADDR_SPACE_IO 0x1
+#define OF_PCI_ADDR_SPACE_MEM32 0x2
+#define OF_PCI_ADDR_SPACE_MEM64 0x3
+
+#define OF_PCI_ADDR_FIELD_NONRELOC BIT(31)
+#define OF_PCI_ADDR_FIELD_SS GENMASK(25, 24)
+#define OF_PCI_ADDR_FIELD_PREFETCH BIT(30)
+#define OF_PCI_ADDR_FIELD_BUS GENMASK(23, 16)
+#define OF_PCI_ADDR_FIELD_DEV GENMASK(15, 11)
+#define OF_PCI_ADDR_FIELD_FUNC GENMASK(10, 8)
+#define OF_PCI_ADDR_FIELD_REG GENMASK(7, 0)
+
+#define OF_PCI_ADDR_HI GENMASK_ULL(63, 32)
+#define OF_PCI_ADDR_LO GENMASK_ULL(31, 0)
+#define OF_PCI_SIZE_HI GENMASK_ULL(63, 32)
+#define OF_PCI_SIZE_LO GENMASK_ULL(31, 0)
+
+#define OF_PCI_ADDRESS_CELLS 3
+#define OF_PCI_SIZE_CELLS 2
+
+enum of_pci_prop_compatible {
+ PROP_COMPAT_PCI_VVVV_DDDD,
+ PROP_COMPAT_PCICLASS_CCSSPP,
+ PROP_COMPAT_PCICLASS_CCSS,
+ PROP_COMPAT_NUM,
+};
+
+static int of_pci_prop_device_type(struct pci_dev *pdev,
+ struct of_changeset *ocs,
+ struct device_node *np)
+{
+ return of_changeset_add_prop_string(ocs, np, "device_type", "pci");
+}
+
+static int of_pci_prop_address_cells(struct pci_dev *pdev,
+ struct of_changeset *ocs,
+ struct device_node *np)
+{
+ return of_changeset_add_prop_u32(ocs, np, "#address_cells",
+ OF_PCI_ADDRESS_CELLS);
+}
+
+static int of_pci_prop_size_cells(struct pci_dev *pdev,
+ struct of_changeset *ocs,
+ struct device_node *np)
+{
+ return of_changeset_add_prop_u32(ocs, np, "#size_cells",
+ OF_PCI_SIZE_CELLS);
+}
+
+static int of_pci_set_addr_flags(struct resource *res, u32 *addr_hi)
+{
+ u32 ss;
+
+ if (res->flags & IORESOURCE_IO)
+ ss = OF_PCI_ADDR_SPACE_IO;
+ else if (res->flags & IORESOURCE_MEM_64)
+ ss = OF_PCI_ADDR_SPACE_MEM64;
+ else if (res->flags & IORESOURCE_MEM)
+ ss = OF_PCI_ADDR_SPACE_MEM32;
+ else
+ return -EINVAL;
+
+ *addr_hi &= ~(OF_PCI_ADDR_FIELD_SS | OF_PCI_ADDR_FIELD_PREFETCH);
+ if (res->flags & IORESOURCE_PREFETCH)
+ *addr_hi |= OF_PCI_ADDR_FIELD_PREFETCH;
+
+ *addr_hi |= ss;
+
+ return 0;
+}
+
+static int of_pci_prop_ranges(struct pci_dev *pdev, struct of_changeset *ocs,
+ struct device_node *np)
+{
+ struct of_pci_range rp[PCI_BRIDGE_RESOURCE_NUM];
+ struct resource *res;
+ int i = 0, j, ret;
+ u64 val64;
+ u32 val;
+
+ res = &pdev->resource[PCI_BRIDGE_RESOURCES];
+ for (j = 0; j < PCI_BRIDGE_RESOURCE_NUM; j++) {
+ if (!resource_size(&res[j]))
+ continue;
+
+ val = OF_PCI_ADDR_FIELD_NONRELOC;
+ if (of_pci_set_addr_flags(&res[j], &val))
+ continue;
+
+ rp[i].parent_addr_hi = cpu_to_be32(val);
+
+ val64 = res[j].start;
+ rp[i].parent_addr_mid =
+ cpu_to_be32(FIELD_GET(OF_PCI_ADDR_HI, val64));
+ rp[i].parent_addr_lo =
+ cpu_to_be32(FIELD_GET(OF_PCI_ADDR_LO, val64));
+
+ val64 = resource_size(&res[j]);
+ rp[i].size_hi = cpu_to_be32(FIELD_GET(OF_PCI_SIZE_HI, val64));
+ rp[i].size_lo = cpu_to_be32(FIELD_GET(OF_PCI_SIZE_LO, val64));
+
+ rp[i].child_addr_hi = rp[i].parent_addr_hi;
+ rp[i].child_addr_mid = rp[i].parent_addr_mid;
+ rp[i].child_addr_lo = rp[i].parent_addr_lo;
+ i++;
+ }
+
+ ret = of_changeset_add_prop_u32_array(ocs, np, "ranges", (u32 *)rp,
+ i * sizeof(*rp) / sizeof(u32));
+
+ return ret;
+}
+
+static int of_pci_prop_reg(struct pci_dev *pdev, struct of_changeset *ocs,
+ struct device_node *np)
+{
+ struct of_pci_addr_pair *reg;
+ int i = 1, resno, ret = 0;
+ u32 reg_val, base_addr;
+ resource_size_t sz;
+
+ reg = kzalloc(sizeof(*reg) * (PCI_STD_NUM_BARS + 1), GFP_KERNEL);
+ if (!reg)
+ return -ENOMEM;
+
+ reg_val = FIELD_PREP(OF_PCI_ADDR_FIELD_SS, OF_PCI_ADDR_SPACE_CONFIG) |
+ FIELD_PREP(OF_PCI_ADDR_FIELD_BUS, pdev->bus->number) |
+ FIELD_PREP(OF_PCI_ADDR_FIELD_DEV, PCI_SLOT(pdev->devfn)) |
+ FIELD_PREP(OF_PCI_ADDR_FIELD_FUNC, PCI_FUNC(pdev->devfn));
+ reg[0].phys_hi = cpu_to_be32(reg_val);
+
+ base_addr = PCI_BASE_ADDRESS_0;
+ for (resno = PCI_STD_RESOURCES; resno <= PCI_STD_RESOURCE_END;
+ resno++, base_addr += 4) {
+ sz = pci_resource_len(pdev, resno);
+ if (!sz)
+ continue;
+
+ ret = of_pci_set_addr_flags(&pdev->resource[resno], ®_val);
+ if (!ret)
+ continue;
+
+ reg_val &= ~OF_PCI_ADDR_FIELD_REG;
+ reg_val |= FIELD_PREP(OF_PCI_ADDR_FIELD_REG, base_addr);
+ reg[i].phys_hi = cpu_to_be32(reg_val);
+ reg[i].size_hi = cpu_to_be32(FIELD_GET(OF_PCI_SIZE_HI, sz));
+ reg[i].size_lo = cpu_to_be32(FIELD_GET(OF_PCI_SIZE_LO, sz));
+ i++;
+ }
+
+ ret = of_changeset_add_prop_u32_array(ocs, np, "reg", (u32 *)reg,
+ i * sizeof(*reg) / sizeof(u32));
+ kfree(reg);
+
+ return ret;
+}
+
+static int of_pci_prop_compatible(struct pci_dev *pdev,
+ struct of_changeset *ocs,
+ struct device_node *np)
+{
+ const char *compat_strs[PROP_COMPAT_NUM] = { 0 };
+ int i, ret;
+
+ compat_strs[PROP_COMPAT_PCI_VVVV_DDDD] =
+ kasprintf(GFP_KERNEL, "pci%x,%x", pdev->vendor, pdev->device);
+ compat_strs[PROP_COMPAT_PCICLASS_CCSSPP] =
+ kasprintf(GFP_KERNEL, "pciclass,%06x", pdev->class);
+ compat_strs[PROP_COMPAT_PCICLASS_CCSS] =
+ kasprintf(GFP_KERNEL, "pciclass,%04x", pdev->class >> 8);
+
+ ret = of_changeset_add_prop_string_array(ocs, np, "compatible",
+ compat_strs, PROP_COMPAT_NUM);
+ for (i = 0; i < PROP_COMPAT_NUM; i++)
+ kfree(compat_strs[i]);
+
+ return ret;
+}
+
+static int (*of_pci_endpoint_props[])(struct pci_dev *pdev,
+ struct of_changeset *ocs,
+ struct device_node *np) = {
+ of_pci_prop_reg,
+ of_pci_prop_compatible,
+ NULL
+};
+
+static int (*of_pci_bridge_props[])(struct pci_dev *pdev,
+ struct of_changeset *ocs,
+ struct device_node *np) = {
+ of_pci_prop_device_type,
+ of_pci_prop_address_cells,
+ of_pci_prop_size_cells,
+ of_pci_prop_ranges,
+ of_pci_prop_reg,
+ of_pci_prop_compatible,
+ NULL
+};
+
+int of_pci_add_properties(struct pci_dev *pdev, struct of_changeset *ocs,
+ struct device_node *np)
+{
+ int (**prop_func)(struct pci_dev *pdev, struct of_changeset *ocs,
+ struct device_node *np);
+ int i, ret;
+
+ if (pci_is_bridge(pdev))
+ prop_func = of_pci_bridge_props;
+ else
+ prop_func = of_pci_endpoint_props;
+
+ for (i = 0; prop_func[i]; i++) {
+ ret = prop_func[i](pdev, ocs, np);
+ if (ret) {
+ /*
+ * The added properties will be released when the
+ * changeset is destroyed.
+ */
+ return ret;
+ }
+ }
+
+ return 0;
+}
diff --git a/drivers/pci/pci-driver.c b/drivers/pci/pci-driver.c
index 49238ddd39ee..1540c4c9a770 100644
--- a/drivers/pci/pci-driver.c
+++ b/drivers/pci/pci-driver.c
@@ -1628,7 +1628,8 @@ static int pci_dma_configure(struct device *dev)
bridge = pci_get_host_bridge_device(to_pci_dev(dev));
if (IS_ENABLED(CONFIG_OF) && bridge->parent &&
- bridge->parent->of_node) {
+ bridge->parent->of_node &&
+ !of_node_check_flag(bridge->parent->of_node, OF_DYNAMIC)) {
ret = of_dma_configure(dev, bridge->parent->of_node, true);
} else if (has_acpi_companion(bridge)) {
struct acpi_device *adev = to_acpi_device_node(bridge->fwnode);
diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
index 785f31086313..bd81dc4ca04f 100644
--- a/drivers/pci/pci.h
+++ b/drivers/pci/pci.h
@@ -678,6 +678,25 @@ static inline int devm_of_pci_bridge_init(struct device *dev, struct pci_host_br
#endif /* CONFIG_OF */
+struct of_changeset;
+
+#ifdef CONFIG_PCI_DYNAMIC_OF_NODES
+void of_pci_make_dev_node(struct pci_dev *pdev);
+void of_pci_remove_node(struct pci_dev *pdev);
+int of_pci_add_properties(struct pci_dev *pdev, struct of_changeset *ocs,
+ struct device_node *np);
+#else
+static inline void
+of_pci_make_dev_node(struct pci_dev *pdev)
+{
+}
+
+static inline void
+of_pci_remove_node(struct pci_dev *pdev)
+{
+}
+#endif /* CONFIG_PCI_DYNAMIC_OF_NODES */
+
#ifdef CONFIG_PCIEAER
void pci_no_aer(void);
void pci_aer_init(struct pci_dev *dev);
diff --git a/drivers/pci/remove.c b/drivers/pci/remove.c
index 4c54c75050dc..0eaa9d9a3609 100644
--- a/drivers/pci/remove.c
+++ b/drivers/pci/remove.c
@@ -23,6 +23,7 @@ static void pci_stop_dev(struct pci_dev *dev)
device_release_driver(&dev->dev);
pci_proc_detach_device(dev);
pci_remove_sysfs_dev_files(dev);
+ of_pci_remove_node(dev);
pci_dev_assign_added(dev, false);
}
--
2.17.1
On Mon, Nov 21, 2022 at 08:43:03AM -0800, Lizhi Hou wrote:
> The PCI endpoint device such as Xilinx Alveo PCI card maps the register
> spaces from multiple hardware peripherals to its PCI BAR. Normally,
> the PCI core discovers devices and BARs using the PCI enumeration process.
> There is no infrastructure to discover the hardware peripherals that are
> present in a PCI device, and which can be accessed through the PCI BARs.
>
> For Alveo PCI card, the card firmware provides a flattened device tree to
> describe the hardware peripherals on its BARs. The Alveo card driver can
> load this flattened device tree and leverage device tree framework to
> generate platform devices for the hardware peripherals eventually.
>
> Apparently, the device tree framework requires a device tree node for the
> PCI device. Thus, it can generate the device tree nodes for hardware
> peripherals underneath. Because PCI is self discoverable bus, there might
> not be a device tree node created for PCI devices. This patch is to add
> support to generate device tree node for PCI devices.
>
> Added a kernel option. When the option is turned on, the kernel will
> generate device tree nodes for PCI bridges unconditionally.
>
> Initially, the basic properties are added for the dynamically generated
> device tree nodes.
>
> Signed-off-by: Lizhi Hou <[email protected]>
> Signed-off-by: Sonal Santan <[email protected]>
> Signed-off-by: Max Zhen <[email protected]>
> Reviewed-by: Brian Xu <[email protected]>
> ---
> drivers/pci/Kconfig | 12 ++
> drivers/pci/Makefile | 1 +
> drivers/pci/bus.c | 2 +
> drivers/pci/msi/irqdomain.c | 6 +-
> drivers/pci/of.c | 71 ++++++++++
> drivers/pci/of_property.c | 256 ++++++++++++++++++++++++++++++++++++
> drivers/pci/pci-driver.c | 3 +-
> drivers/pci/pci.h | 19 +++
> drivers/pci/remove.c | 1 +
> 9 files changed, 368 insertions(+), 3 deletions(-)
> create mode 100644 drivers/pci/of_property.c
>
> diff --git a/drivers/pci/Kconfig b/drivers/pci/Kconfig
> index 55c028af4bd9..126c31b79718 100644
> --- a/drivers/pci/Kconfig
> +++ b/drivers/pci/Kconfig
> @@ -198,6 +198,18 @@ config PCI_HYPERV
> The PCI device frontend driver allows the kernel to import arbitrary
> PCI devices from a PCI backend to support PCI driver domains.
>
> +config PCI_DYNAMIC_OF_NODES
> + bool "Device tree node for PCI devices"
Create Devicetree nodes for PCI devices
But as I've said before, making this a config option doesn't really work
except as something to experiment with. Once you add your driver and
want to do a 'select PCI_DYNAMIC_OF_NODES', you've affected everyone
else.
> + depends on OF
> + select OF_DYNAMIC
> + help
> + This option enables support for generating device tree nodes for some
> + PCI devices. Thus, the driver of this kind can load and overlay
> + flattened device tree for its downstream devices.
> +
> + Once this option is selected, the device tree nodes will be generated
> + for all PCI bridges.
> +
> choice
> prompt "PCI Express hierarchy optimization setting"
> default PCIE_BUS_DEFAULT
> diff --git a/drivers/pci/Makefile b/drivers/pci/Makefile
> index 2680e4c92f0a..cc8b4e01e29d 100644
> --- a/drivers/pci/Makefile
> +++ b/drivers/pci/Makefile
> @@ -32,6 +32,7 @@ obj-$(CONFIG_PCI_P2PDMA) += p2pdma.o
> obj-$(CONFIG_XEN_PCIDEV_FRONTEND) += xen-pcifront.o
> obj-$(CONFIG_VGA_ARB) += vgaarb.o
> obj-$(CONFIG_PCI_DOE) += doe.o
> +obj-$(CONFIG_PCI_DYNAMIC_OF_NODES) += of_property.o
>
> # Endpoint library must be initialized before its users
> obj-$(CONFIG_PCI_ENDPOINT) += endpoint/
> diff --git a/drivers/pci/bus.c b/drivers/pci/bus.c
> index 3cef835b375f..8507cc32b61d 100644
> --- a/drivers/pci/bus.c
> +++ b/drivers/pci/bus.c
> @@ -316,6 +316,8 @@ void pci_bus_add_device(struct pci_dev *dev)
> */
> pcibios_bus_add_device(dev);
> pci_fixup_device(pci_fixup_final, dev);
> + if (pci_is_bridge(dev))
> + of_pci_make_dev_node(dev);
> pci_create_sysfs_dev_files(dev);
> pci_proc_attach_device(dev);
> pci_bridge_d3_update(dev);
> diff --git a/drivers/pci/msi/irqdomain.c b/drivers/pci/msi/irqdomain.c
> index e9cf318e6670..eeaf44169bfd 100644
> --- a/drivers/pci/msi/irqdomain.c
> +++ b/drivers/pci/msi/irqdomain.c
> @@ -230,8 +230,10 @@ u32 pci_msi_domain_get_msi_rid(struct irq_domain *domain, struct pci_dev *pdev)
> pci_for_each_dma_alias(pdev, get_msi_id_cb, &rid);
>
> of_node = irq_domain_get_of_node(domain);
> - rid = of_node ? of_msi_map_id(&pdev->dev, of_node, rid) :
> - iort_msi_map_id(&pdev->dev, rid);
> + if (of_node && !of_node_check_flag(of_node, OF_DYNAMIC))
> + rid = of_msi_map_id(&pdev->dev, of_node, rid);
> + else
> + rid = iort_msi_map_id(&pdev->dev, rid);
I have no idea if this works. It looks kind of broken already if
!of_node calls iort_msi_map_id(). With a DT only system, I would think
we'd always call of_msi_map_id(). Have you tested MSIs?
With a mixed system, I have no idea what happens. That needs to be
understood for MSI, DMA, and interrupts.
>
> return rid;
> }
> diff --git a/drivers/pci/of.c b/drivers/pci/of.c
> index 196834ed44fe..fb60b04f0b93 100644
> --- a/drivers/pci/of.c
> +++ b/drivers/pci/of.c
> @@ -469,6 +469,8 @@ static int of_irq_parse_pci(const struct pci_dev *pdev, struct of_phandle_args *
> } else {
> /* We found a P2P bridge, check if it has a node */
> ppnode = pci_device_to_OF_node(ppdev);
> + if (of_node_check_flag(ppnode, OF_DYNAMIC))
> + ppnode = NULL;
> }
>
> /*
> @@ -599,6 +601,75 @@ int devm_of_pci_bridge_init(struct device *dev, struct pci_host_bridge *bridge)
> return pci_parse_request_of_pci_ranges(dev, bridge);
> }
>
> +#if IS_ENABLED(CONFIG_PCI_DYNAMIC_OF_NODES)
> +
> +void of_pci_remove_node(struct pci_dev *pdev)
> +{
> + struct device_node *dt_node;
node or np are the typical names.
> +
> + dt_node = pci_device_to_OF_node(pdev);
> + if (!dt_node || !of_node_check_flag(dt_node, OF_DYNAMIC))
> + return;
> + pdev->dev.of_node = NULL;
> +
> + of_destroy_node(dt_node);
> +}
> +
> +void of_pci_make_dev_node(struct pci_dev *pdev)
> +{
> + struct device_node *parent, *dt_node = NULL;
> + const char *pci_type = "dev";
> + struct of_changeset *cset;
> + const char *full_name;
> + int ret;
> +
> + /*
> + * If there is already a device tree node linked to this device,
> + * return immediately.
> + */
> + if (pci_device_to_OF_node(pdev))
> + return;
> +
> + /* Check if there is device tree node for parent device */
> + if (!pdev->bus->self)
> + parent = pdev->bus->dev.of_node;
> + else
> + parent = pdev->bus->self->dev.of_node;
> + if (!parent)
> + return;
> +
> + if (pci_is_bridge(pdev))
> + pci_type = "pci";
What's the node name if not a bridge? I don't see how that would work.
It should depend on the device class if it has one.
> +
> + full_name = kasprintf(GFP_KERNEL, "%pOF/%s@%x,%x", parent, pci_type,
> + PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
'full_name' in the DT code is historical. It's not storing the full path
now. It's just 'name' with the unit-address. So why does
of_create_node() need the full path?
> + if (!full_name)
> + goto failed;
> +
> + dt_node = of_create_node(parent, full_name, &cset);
> + if (!dt_node)
> + goto failed;
> + kfree(full_name);
> +
> + ret = of_pci_add_properties(pdev, cset, dt_node);
> + if (ret)
> + goto failed;
> +
> + ret = of_changeset_apply(cset);
> + if (ret)
> + goto failed;
> +
> + pdev->dev.of_node = dt_node;
> +
> + return;
> +
> +failed:
> + if (dt_node)
> + of_destroy_node(dt_node);
> + kfree(full_name);
> +}
> +#endif
> +
> #endif /* CONFIG_PCI */
>
> /**
> diff --git a/drivers/pci/of_property.c b/drivers/pci/of_property.c
> new file mode 100644
> index 000000000000..cc66fa7517e0
> --- /dev/null
> +++ b/drivers/pci/of_property.c
> @@ -0,0 +1,256 @@
> +// SPDX-License-Identifier: GPL-2.0+
GPL-2.0-only
> +/*
> + * Copyright (C) 2022, Advanced Micro Devices, Inc.
> + */
> +
> +#include <linux/pci.h>
> +#include <linux/of.h>
> +#include <linux/bitfield.h>
> +#include <linux/bits.h>
> +#include "pci.h"
> +
> +struct of_pci_addr_pair {
> + __be32 phys_hi;
> + __be32 phys_mid;
> + __be32 phys_lo;
> + __be32 size_hi;
> + __be32 size_lo;
> +};
> +
> +struct of_pci_range {
> + __be32 child_addr_hi;
> + __be32 child_addr_mid;
> + __be32 child_addr_lo;
> + __be32 parent_addr_hi;
> + __be32 parent_addr_mid;
> + __be32 parent_addr_lo;
> + __be32 size_hi;
> + __be32 size_lo;
> +};
> +
> +#define OF_PCI_ADDR_SPACE_CONFIG 0x0
> +#define OF_PCI_ADDR_SPACE_IO 0x1
> +#define OF_PCI_ADDR_SPACE_MEM32 0x2
> +#define OF_PCI_ADDR_SPACE_MEM64 0x3
> +
> +#define OF_PCI_ADDR_FIELD_NONRELOC BIT(31)
> +#define OF_PCI_ADDR_FIELD_SS GENMASK(25, 24)
> +#define OF_PCI_ADDR_FIELD_PREFETCH BIT(30)
> +#define OF_PCI_ADDR_FIELD_BUS GENMASK(23, 16)
> +#define OF_PCI_ADDR_FIELD_DEV GENMASK(15, 11)
> +#define OF_PCI_ADDR_FIELD_FUNC GENMASK(10, 8)
> +#define OF_PCI_ADDR_FIELD_REG GENMASK(7, 0)
> +
> +#define OF_PCI_ADDR_HI GENMASK_ULL(63, 32)
> +#define OF_PCI_ADDR_LO GENMASK_ULL(31, 0)
> +#define OF_PCI_SIZE_HI GENMASK_ULL(63, 32)
> +#define OF_PCI_SIZE_LO GENMASK_ULL(31, 0)
> +
> +#define OF_PCI_ADDRESS_CELLS 3
> +#define OF_PCI_SIZE_CELLS 2
> +
> +enum of_pci_prop_compatible {
> + PROP_COMPAT_PCI_VVVV_DDDD,
> + PROP_COMPAT_PCICLASS_CCSSPP,
> + PROP_COMPAT_PCICLASS_CCSS,
> + PROP_COMPAT_NUM,
> +};
> +
> +static int of_pci_prop_device_type(struct pci_dev *pdev,
> + struct of_changeset *ocs,
> + struct device_node *np)
> +{
> + return of_changeset_add_prop_string(ocs, np, "device_type", "pci");
> +}
> +
> +static int of_pci_prop_address_cells(struct pci_dev *pdev,
> + struct of_changeset *ocs,
> + struct device_node *np)
> +{
> + return of_changeset_add_prop_u32(ocs, np, "#address_cells",
> + OF_PCI_ADDRESS_CELLS);
> +}
> +
> +static int of_pci_prop_size_cells(struct pci_dev *pdev,
> + struct of_changeset *ocs,
> + struct device_node *np)
> +{
> + return of_changeset_add_prop_u32(ocs, np, "#size_cells",
> + OF_PCI_SIZE_CELLS);
> +}
> +
> +static int of_pci_set_addr_flags(struct resource *res, u32 *addr_hi)
> +{
> + u32 ss;
> +
> + if (res->flags & IORESOURCE_IO)
> + ss = OF_PCI_ADDR_SPACE_IO;
> + else if (res->flags & IORESOURCE_MEM_64)
> + ss = OF_PCI_ADDR_SPACE_MEM64;
> + else if (res->flags & IORESOURCE_MEM)
> + ss = OF_PCI_ADDR_SPACE_MEM32;
> + else
> + return -EINVAL;
> +
> + *addr_hi &= ~(OF_PCI_ADDR_FIELD_SS | OF_PCI_ADDR_FIELD_PREFETCH);
> + if (res->flags & IORESOURCE_PREFETCH)
> + *addr_hi |= OF_PCI_ADDR_FIELD_PREFETCH;
> +
> + *addr_hi |= ss;
> +
> + return 0;
> +}
> +
> +static int of_pci_prop_ranges(struct pci_dev *pdev, struct of_changeset *ocs,
> + struct device_node *np)
> +{
> + struct of_pci_range rp[PCI_BRIDGE_RESOURCE_NUM];
> + struct resource *res;
> + int i = 0, j, ret;
> + u64 val64;
> + u32 val;
> +
> + res = &pdev->resource[PCI_BRIDGE_RESOURCES];
> + for (j = 0; j < PCI_BRIDGE_RESOURCE_NUM; j++) {
> + if (!resource_size(&res[j]))
> + continue;
> +
> + val = OF_PCI_ADDR_FIELD_NONRELOC;
> + if (of_pci_set_addr_flags(&res[j], &val))
> + continue;
> +
> + rp[i].parent_addr_hi = cpu_to_be32(val);
> +
> + val64 = res[j].start;
> + rp[i].parent_addr_mid =
> + cpu_to_be32(FIELD_GET(OF_PCI_ADDR_HI, val64));
> + rp[i].parent_addr_lo =
> + cpu_to_be32(FIELD_GET(OF_PCI_ADDR_LO, val64));
cpu_to_be64(res[j].start) would be simpler, but then you'll need
unaligned accessors.
This all could use some refactoring as you write 3 cell PCI addreses in
several places. Something like:
of_pci_set_address(__be32 *prop, u64 addr, u32 flags)
> +
> + val64 = resource_size(&res[j]);
> + rp[i].size_hi = cpu_to_be32(FIELD_GET(OF_PCI_SIZE_HI, val64));
> + rp[i].size_lo = cpu_to_be32(FIELD_GET(OF_PCI_SIZE_LO, val64));
> +
> + rp[i].child_addr_hi = rp[i].parent_addr_hi;
> + rp[i].child_addr_mid = rp[i].parent_addr_mid;
> + rp[i].child_addr_lo = rp[i].parent_addr_lo;
> + i++;
> + }
> +
> + ret = of_changeset_add_prop_u32_array(ocs, np, "ranges", (u32 *)rp,
> + i * sizeof(*rp) / sizeof(u32));
> +
> + return ret;
> +}
> +
> +static int of_pci_prop_reg(struct pci_dev *pdev, struct of_changeset *ocs,
> + struct device_node *np)
> +{
> + struct of_pci_addr_pair *reg;
> + int i = 1, resno, ret = 0;
> + u32 reg_val, base_addr;
> + resource_size_t sz;
> +
> + reg = kzalloc(sizeof(*reg) * (PCI_STD_NUM_BARS + 1), GFP_KERNEL);
> + if (!reg)
> + return -ENOMEM;
> +
> + reg_val = FIELD_PREP(OF_PCI_ADDR_FIELD_SS, OF_PCI_ADDR_SPACE_CONFIG) |
> + FIELD_PREP(OF_PCI_ADDR_FIELD_BUS, pdev->bus->number) |
> + FIELD_PREP(OF_PCI_ADDR_FIELD_DEV, PCI_SLOT(pdev->devfn)) |
> + FIELD_PREP(OF_PCI_ADDR_FIELD_FUNC, PCI_FUNC(pdev->devfn));
> + reg[0].phys_hi = cpu_to_be32(reg_val);
> +
> + base_addr = PCI_BASE_ADDRESS_0;
> + for (resno = PCI_STD_RESOURCES; resno <= PCI_STD_RESOURCE_END;
> + resno++, base_addr += 4) {
> + sz = pci_resource_len(pdev, resno);
> + if (!sz)
> + continue;
> +
> + ret = of_pci_set_addr_flags(&pdev->resource[resno], ®_val);
> + if (!ret)
> + continue;
> +
> + reg_val &= ~OF_PCI_ADDR_FIELD_REG;
> + reg_val |= FIELD_PREP(OF_PCI_ADDR_FIELD_REG, base_addr);
> + reg[i].phys_hi = cpu_to_be32(reg_val);
> + reg[i].size_hi = cpu_to_be32(FIELD_GET(OF_PCI_SIZE_HI, sz));
> + reg[i].size_lo = cpu_to_be32(FIELD_GET(OF_PCI_SIZE_LO, sz));
> + i++;
> + }
> +
> + ret = of_changeset_add_prop_u32_array(ocs, np, "reg", (u32 *)reg,
> + i * sizeof(*reg) / sizeof(u32));
> + kfree(reg);
> +
> + return ret;
> +}
> +
> +static int of_pci_prop_compatible(struct pci_dev *pdev,
> + struct of_changeset *ocs,
> + struct device_node *np)
> +{
> + const char *compat_strs[PROP_COMPAT_NUM] = { 0 };
> + int i, ret;
> +
> + compat_strs[PROP_COMPAT_PCI_VVVV_DDDD] =
> + kasprintf(GFP_KERNEL, "pci%x,%x", pdev->vendor, pdev->device);
> + compat_strs[PROP_COMPAT_PCICLASS_CCSSPP] =
> + kasprintf(GFP_KERNEL, "pciclass,%06x", pdev->class);
> + compat_strs[PROP_COMPAT_PCICLASS_CCSS] =
> + kasprintf(GFP_KERNEL, "pciclass,%04x", pdev->class >> 8);
> +
> + ret = of_changeset_add_prop_string_array(ocs, np, "compatible",
> + compat_strs, PROP_COMPAT_NUM);
> + for (i = 0; i < PROP_COMPAT_NUM; i++)
> + kfree(compat_strs[i]);
> +
> + return ret;
> +}
> +
> +static int (*of_pci_endpoint_props[])(struct pci_dev *pdev,
> + struct of_changeset *ocs,
> + struct device_node *np) = {
> + of_pci_prop_reg,
> + of_pci_prop_compatible,
> + NULL
> +};
> +
> +static int (*of_pci_bridge_props[])(struct pci_dev *pdev,
> + struct of_changeset *ocs,
> + struct device_node *np) = {
> + of_pci_prop_device_type,
> + of_pci_prop_address_cells,
> + of_pci_prop_size_cells,
> + of_pci_prop_ranges,
> + of_pci_prop_reg,
> + of_pci_prop_compatible,
> + NULL
> +};
> +
> +int of_pci_add_properties(struct pci_dev *pdev, struct of_changeset *ocs,
> + struct device_node *np)
> +{
> + int (**prop_func)(struct pci_dev *pdev, struct of_changeset *ocs,
> + struct device_node *np);
> + int i, ret;
> +
> + if (pci_is_bridge(pdev))
> + prop_func = of_pci_bridge_props;
Skip all this indirection and just make everything direct function
calls.
> + else
> + prop_func = of_pci_endpoint_props;
> +
> + for (i = 0; prop_func[i]; i++) {
> + ret = prop_func[i](pdev, ocs, np);
> + if (ret) {
> + /*
> + * The added properties will be released when the
> + * changeset is destroyed.
> + */
> + return ret;
> + }
> + }
> +
> + return 0;
> +}
> diff --git a/drivers/pci/pci-driver.c b/drivers/pci/pci-driver.c
> index 49238ddd39ee..1540c4c9a770 100644
> --- a/drivers/pci/pci-driver.c
> +++ b/drivers/pci/pci-driver.c
> @@ -1628,7 +1628,8 @@ static int pci_dma_configure(struct device *dev)
> bridge = pci_get_host_bridge_device(to_pci_dev(dev));
>
> if (IS_ENABLED(CONFIG_OF) && bridge->parent &&
> - bridge->parent->of_node) {
> + bridge->parent->of_node &&
> + !of_node_check_flag(bridge->parent->of_node, OF_DYNAMIC)) {
> ret = of_dma_configure(dev, bridge->parent->of_node, true);
> } else if (has_acpi_companion(bridge)) {
> struct acpi_device *adev = to_acpi_device_node(bridge->fwnode);
> diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
> index 785f31086313..bd81dc4ca04f 100644
> --- a/drivers/pci/pci.h
> +++ b/drivers/pci/pci.h
> @@ -678,6 +678,25 @@ static inline int devm_of_pci_bridge_init(struct device *dev, struct pci_host_br
>
> #endif /* CONFIG_OF */
>
> +struct of_changeset;
> +
> +#ifdef CONFIG_PCI_DYNAMIC_OF_NODES
> +void of_pci_make_dev_node(struct pci_dev *pdev);
> +void of_pci_remove_node(struct pci_dev *pdev);
> +int of_pci_add_properties(struct pci_dev *pdev, struct of_changeset *ocs,
> + struct device_node *np);
> +#else
> +static inline void
> +of_pci_make_dev_node(struct pci_dev *pdev)
> +{
> +}
> +
> +static inline void
> +of_pci_remove_node(struct pci_dev *pdev)
> +{
> +}
> +#endif /* CONFIG_PCI_DYNAMIC_OF_NODES */
> +
> #ifdef CONFIG_PCIEAER
> void pci_no_aer(void);
> void pci_aer_init(struct pci_dev *dev);
> diff --git a/drivers/pci/remove.c b/drivers/pci/remove.c
> index 4c54c75050dc..0eaa9d9a3609 100644
> --- a/drivers/pci/remove.c
> +++ b/drivers/pci/remove.c
> @@ -23,6 +23,7 @@ static void pci_stop_dev(struct pci_dev *dev)
> device_release_driver(&dev->dev);
> pci_proc_detach_device(dev);
> pci_remove_sysfs_dev_files(dev);
> + of_pci_remove_node(dev);
>
> pci_dev_assign_added(dev, false);
> }
> --
> 2.17.1
>
>
On 12/1/22 13:12, Rob Herring wrote:
> On Mon, Nov 21, 2022 at 08:43:03AM -0800, Lizhi Hou wrote:
>> The PCI endpoint device such as Xilinx Alveo PCI card maps the register
>> spaces from multiple hardware peripherals to its PCI BAR. Normally,
>> the PCI core discovers devices and BARs using the PCI enumeration process.
>> There is no infrastructure to discover the hardware peripherals that are
>> present in a PCI device, and which can be accessed through the PCI BARs.
>>
>> For Alveo PCI card, the card firmware provides a flattened device tree to
>> describe the hardware peripherals on its BARs. The Alveo card driver can
>> load this flattened device tree and leverage device tree framework to
>> generate platform devices for the hardware peripherals eventually.
>>
>> Apparently, the device tree framework requires a device tree node for the
>> PCI device. Thus, it can generate the device tree nodes for hardware
>> peripherals underneath. Because PCI is self discoverable bus, there might
>> not be a device tree node created for PCI devices. This patch is to add
>> support to generate device tree node for PCI devices.
>>
>> Added a kernel option. When the option is turned on, the kernel will
>> generate device tree nodes for PCI bridges unconditionally.
>>
>> Initially, the basic properties are added for the dynamically generated
>> device tree nodes.
>>
>> Signed-off-by: Lizhi Hou <[email protected]>
>> Signed-off-by: Sonal Santan <[email protected]>
>> Signed-off-by: Max Zhen <[email protected]>
>> Reviewed-by: Brian Xu <[email protected]>
>> ---
>> drivers/pci/Kconfig | 12 ++
>> drivers/pci/Makefile | 1 +
>> drivers/pci/bus.c | 2 +
>> drivers/pci/msi/irqdomain.c | 6 +-
>> drivers/pci/of.c | 71 ++++++++++
>> drivers/pci/of_property.c | 256 ++++++++++++++++++++++++++++++++++++
>> drivers/pci/pci-driver.c | 3 +-
>> drivers/pci/pci.h | 19 +++
>> drivers/pci/remove.c | 1 +
>> 9 files changed, 368 insertions(+), 3 deletions(-)
>> create mode 100644 drivers/pci/of_property.c
>>
>> diff --git a/drivers/pci/Kconfig b/drivers/pci/Kconfig
>> index 55c028af4bd9..126c31b79718 100644
>> --- a/drivers/pci/Kconfig
>> +++ b/drivers/pci/Kconfig
>> @@ -198,6 +198,18 @@ config PCI_HYPERV
>> The PCI device frontend driver allows the kernel to import arbitrary
>> PCI devices from a PCI backend to support PCI driver domains.
>>
>> +config PCI_DYNAMIC_OF_NODES
>> + bool "Device tree node for PCI devices"
> Create Devicetree nodes for PCI devices
Sure.
>
> But as I've said before, making this a config option doesn't really work
> except as something to experiment with. Once you add your driver and
> want to do a 'select PCI_DYNAMIC_OF_NODES', you've affected everyone
> else.
Do you mean we should remove PCI_DYNAMIC_OF_NODES and make
creating dynamic tree node default?
Based on the previous discussions, the approach I am implementing
is to create device tree nodes for all PCI bridges and devices defined
by pci quirks. I believe a PCI endpoint which is not defined by PCI quirks
should not to be affected because there is no device tree node is created
for it.
Are you fine with this approach?
>
>> + depends on OF
>> + select OF_DYNAMIC
>> + help
>> + This option enables support for generating device tree nodes for some
>> + PCI devices. Thus, the driver of this kind can load and overlay
>> + flattened device tree for its downstream devices.
>> +
>> + Once this option is selected, the device tree nodes will be generated
>> + for all PCI bridges.
>> +
>> choice
>> prompt "PCI Express hierarchy optimization setting"
>> default PCIE_BUS_DEFAULT
>> diff --git a/drivers/pci/Makefile b/drivers/pci/Makefile
>> index 2680e4c92f0a..cc8b4e01e29d 100644
>> --- a/drivers/pci/Makefile
>> +++ b/drivers/pci/Makefile
>> @@ -32,6 +32,7 @@ obj-$(CONFIG_PCI_P2PDMA) += p2pdma.o
>> obj-$(CONFIG_XEN_PCIDEV_FRONTEND) += xen-pcifront.o
>> obj-$(CONFIG_VGA_ARB) += vgaarb.o
>> obj-$(CONFIG_PCI_DOE) += doe.o
>> +obj-$(CONFIG_PCI_DYNAMIC_OF_NODES) += of_property.o
>>
>> # Endpoint library must be initialized before its users
>> obj-$(CONFIG_PCI_ENDPOINT) += endpoint/
>> diff --git a/drivers/pci/bus.c b/drivers/pci/bus.c
>> index 3cef835b375f..8507cc32b61d 100644
>> --- a/drivers/pci/bus.c
>> +++ b/drivers/pci/bus.c
>> @@ -316,6 +316,8 @@ void pci_bus_add_device(struct pci_dev *dev)
>> */
>> pcibios_bus_add_device(dev);
>> pci_fixup_device(pci_fixup_final, dev);
>> + if (pci_is_bridge(dev))
>> + of_pci_make_dev_node(dev);
>> pci_create_sysfs_dev_files(dev);
>> pci_proc_attach_device(dev);
>> pci_bridge_d3_update(dev);
>> diff --git a/drivers/pci/msi/irqdomain.c b/drivers/pci/msi/irqdomain.c
>> index e9cf318e6670..eeaf44169bfd 100644
>> --- a/drivers/pci/msi/irqdomain.c
>> +++ b/drivers/pci/msi/irqdomain.c
>> @@ -230,8 +230,10 @@ u32 pci_msi_domain_get_msi_rid(struct irq_domain *domain, struct pci_dev *pdev)
>> pci_for_each_dma_alias(pdev, get_msi_id_cb, &rid);
>>
>> of_node = irq_domain_get_of_node(domain);
>> - rid = of_node ? of_msi_map_id(&pdev->dev, of_node, rid) :
>> - iort_msi_map_id(&pdev->dev, rid);
>> + if (of_node && !of_node_check_flag(of_node, OF_DYNAMIC))
>> + rid = of_msi_map_id(&pdev->dev, of_node, rid);
>> + else
>> + rid = iort_msi_map_id(&pdev->dev, rid);
> I have no idea if this works. It looks kind of broken already if
> !of_node calls iort_msi_map_id(). With a DT only system, I would think
> we'd always call of_msi_map_id(). Have you tested MSIs?
>
> With a mixed system, I have no idea what happens. That needs to be
> understood for MSI, DMA, and interrupts.
Yes, I tested MSI in VM.
# cat
/sys/devices/platform/3f000000.pcie/pci0000:00/0000:00:02.0/0000:09:00.0/0000:0a:00.0/msi_irqs/29
msi
# cat /proc/interrupts | grep 29
29: 1 0 MSI 5242880 Edge pciehp
The idea is to preserve the current behaviror.
current: PCI device does not have dt node, thus iort_msi_map_id()
is called.
modified code: PCI device has dt node but with OF_DYNAMIC flag,
thus iort_msi_map_id() is called.
I was planning to take on of_msi_map_id() for dynamically generated dt node
in future when we see a real use case?
>
>>
>> return rid;
>> }
>> diff --git a/drivers/pci/of.c b/drivers/pci/of.c
>> index 196834ed44fe..fb60b04f0b93 100644
>> --- a/drivers/pci/of.c
>> +++ b/drivers/pci/of.c
>> @@ -469,6 +469,8 @@ static int of_irq_parse_pci(const struct pci_dev *pdev, struct of_phandle_args *
>> } else {
>> /* We found a P2P bridge, check if it has a node */
>> ppnode = pci_device_to_OF_node(ppdev);
>> + if (of_node_check_flag(ppnode, OF_DYNAMIC))
>> + ppnode = NULL;
>> }
>>
>> /*
>> @@ -599,6 +601,75 @@ int devm_of_pci_bridge_init(struct device *dev, struct pci_host_bridge *bridge)
>> return pci_parse_request_of_pci_ranges(dev, bridge);
>> }
>>
>> +#if IS_ENABLED(CONFIG_PCI_DYNAMIC_OF_NODES)
>> +
>> +void of_pci_remove_node(struct pci_dev *pdev)
>> +{
>> + struct device_node *dt_node;
> node or np are the typical names.
Will fix this.
>
>> +
>> + dt_node = pci_device_to_OF_node(pdev);
>> + if (!dt_node || !of_node_check_flag(dt_node, OF_DYNAMIC))
>> + return;
>> + pdev->dev.of_node = NULL;
>> +
>> + of_destroy_node(dt_node);
>> +}
>> +
>> +void of_pci_make_dev_node(struct pci_dev *pdev)
>> +{
>> + struct device_node *parent, *dt_node = NULL;
>> + const char *pci_type = "dev";
>> + struct of_changeset *cset;
>> + const char *full_name;
>> + int ret;
>> +
>> + /*
>> + * If there is already a device tree node linked to this device,
>> + * return immediately.
>> + */
>> + if (pci_device_to_OF_node(pdev))
>> + return;
>> +
>> + /* Check if there is device tree node for parent device */
>> + if (!pdev->bus->self)
>> + parent = pdev->bus->dev.of_node;
>> + else
>> + parent = pdev->bus->self->dev.of_node;
>> + if (!parent)
>> + return;
>> +
>> + if (pci_is_bridge(pdev))
>> + pci_type = "pci";
> What's the node name if not a bridge? I don't see how that would work.
>
> It should depend on the device class if it has one.
pci_type is initialized with "dev"
+ const char *pci_type = "dev";
Do you mean I should use class instead of pci_is_bridge()?
if ((pdev->class >> 24) == PCI_BASE_CLASS_BRIDGE)
pci_type = "pci";
>
>> +
>> + full_name = kasprintf(GFP_KERNEL, "%pOF/%s@%x,%x", parent, pci_type,
>> + PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
> 'full_name' in the DT code is historical. It's not storing the full path
> now. It's just 'name' with the unit-address. So why does
> of_create_node() need the full path?
ok. I did not know full_name is historical. I will use 'name' next time.
>> + if (!full_name)
>> + goto failed;
>> +
>> + dt_node = of_create_node(parent, full_name, &cset);
>> + if (!dt_node)
>> + goto failed;
>> + kfree(full_name);
>> +
>> + ret = of_pci_add_properties(pdev, cset, dt_node);
>> + if (ret)
>> + goto failed;
>> +
>> + ret = of_changeset_apply(cset);
>> + if (ret)
>> + goto failed;
>> +
>> + pdev->dev.of_node = dt_node;
>> +
>> + return;
>> +
>> +failed:
>> + if (dt_node)
>> + of_destroy_node(dt_node);
>> + kfree(full_name);
>> +}
>> +#endif
>> +
>> #endif /* CONFIG_PCI */
>>
>> /**
>> diff --git a/drivers/pci/of_property.c b/drivers/pci/of_property.c
>> new file mode 100644
>> index 000000000000..cc66fa7517e0
>> --- /dev/null
>> +++ b/drivers/pci/of_property.c
>> @@ -0,0 +1,256 @@
>> +// SPDX-License-Identifier: GPL-2.0+
> GPL-2.0-only
Got it. Thanks.
>
>> +/*
>> + * Copyright (C) 2022, Advanced Micro Devices, Inc.
>> + */
>> +
>> +#include <linux/pci.h>
>> +#include <linux/of.h>
>> +#include <linux/bitfield.h>
>> +#include <linux/bits.h>
>> +#include "pci.h"
>> +
>> +struct of_pci_addr_pair {
>> + __be32 phys_hi;
>> + __be32 phys_mid;
>> + __be32 phys_lo;
>> + __be32 size_hi;
>> + __be32 size_lo;
>> +};
>> +
>> +struct of_pci_range {
>> + __be32 child_addr_hi;
>> + __be32 child_addr_mid;
>> + __be32 child_addr_lo;
>> + __be32 parent_addr_hi;
>> + __be32 parent_addr_mid;
>> + __be32 parent_addr_lo;
>> + __be32 size_hi;
>> + __be32 size_lo;
>> +};
>> +
>> +#define OF_PCI_ADDR_SPACE_CONFIG 0x0
>> +#define OF_PCI_ADDR_SPACE_IO 0x1
>> +#define OF_PCI_ADDR_SPACE_MEM32 0x2
>> +#define OF_PCI_ADDR_SPACE_MEM64 0x3
>> +
>> +#define OF_PCI_ADDR_FIELD_NONRELOC BIT(31)
>> +#define OF_PCI_ADDR_FIELD_SS GENMASK(25, 24)
>> +#define OF_PCI_ADDR_FIELD_PREFETCH BIT(30)
>> +#define OF_PCI_ADDR_FIELD_BUS GENMASK(23, 16)
>> +#define OF_PCI_ADDR_FIELD_DEV GENMASK(15, 11)
>> +#define OF_PCI_ADDR_FIELD_FUNC GENMASK(10, 8)
>> +#define OF_PCI_ADDR_FIELD_REG GENMASK(7, 0)
>> +
>> +#define OF_PCI_ADDR_HI GENMASK_ULL(63, 32)
>> +#define OF_PCI_ADDR_LO GENMASK_ULL(31, 0)
>> +#define OF_PCI_SIZE_HI GENMASK_ULL(63, 32)
>> +#define OF_PCI_SIZE_LO GENMASK_ULL(31, 0)
>> +
>> +#define OF_PCI_ADDRESS_CELLS 3
>> +#define OF_PCI_SIZE_CELLS 2
>> +
>> +enum of_pci_prop_compatible {
>> + PROP_COMPAT_PCI_VVVV_DDDD,
>> + PROP_COMPAT_PCICLASS_CCSSPP,
>> + PROP_COMPAT_PCICLASS_CCSS,
>> + PROP_COMPAT_NUM,
>> +};
>> +
>> +static int of_pci_prop_device_type(struct pci_dev *pdev,
>> + struct of_changeset *ocs,
>> + struct device_node *np)
>> +{
>> + return of_changeset_add_prop_string(ocs, np, "device_type", "pci");
>> +}
>> +
>> +static int of_pci_prop_address_cells(struct pci_dev *pdev,
>> + struct of_changeset *ocs,
>> + struct device_node *np)
>> +{
>> + return of_changeset_add_prop_u32(ocs, np, "#address_cells",
>> + OF_PCI_ADDRESS_CELLS);
>> +}
>> +
>> +static int of_pci_prop_size_cells(struct pci_dev *pdev,
>> + struct of_changeset *ocs,
>> + struct device_node *np)
>> +{
>> + return of_changeset_add_prop_u32(ocs, np, "#size_cells",
>> + OF_PCI_SIZE_CELLS);
>> +}
>> +
>> +static int of_pci_set_addr_flags(struct resource *res, u32 *addr_hi)
>> +{
>> + u32 ss;
>> +
>> + if (res->flags & IORESOURCE_IO)
>> + ss = OF_PCI_ADDR_SPACE_IO;
>> + else if (res->flags & IORESOURCE_MEM_64)
>> + ss = OF_PCI_ADDR_SPACE_MEM64;
>> + else if (res->flags & IORESOURCE_MEM)
>> + ss = OF_PCI_ADDR_SPACE_MEM32;
>> + else
>> + return -EINVAL;
>> +
>> + *addr_hi &= ~(OF_PCI_ADDR_FIELD_SS | OF_PCI_ADDR_FIELD_PREFETCH);
>> + if (res->flags & IORESOURCE_PREFETCH)
>> + *addr_hi |= OF_PCI_ADDR_FIELD_PREFETCH;
>> +
>> + *addr_hi |= ss;
>> +
>> + return 0;
>> +}
>> +
>> +static int of_pci_prop_ranges(struct pci_dev *pdev, struct of_changeset *ocs,
>> + struct device_node *np)
>> +{
>> + struct of_pci_range rp[PCI_BRIDGE_RESOURCE_NUM];
>> + struct resource *res;
>> + int i = 0, j, ret;
>> + u64 val64;
>> + u32 val;
>> +
>> + res = &pdev->resource[PCI_BRIDGE_RESOURCES];
>> + for (j = 0; j < PCI_BRIDGE_RESOURCE_NUM; j++) {
>> + if (!resource_size(&res[j]))
>> + continue;
>> +
>> + val = OF_PCI_ADDR_FIELD_NONRELOC;
>> + if (of_pci_set_addr_flags(&res[j], &val))
>> + continue;
>> +
>> + rp[i].parent_addr_hi = cpu_to_be32(val);
>> +
>> + val64 = res[j].start;
>> + rp[i].parent_addr_mid =
>> + cpu_to_be32(FIELD_GET(OF_PCI_ADDR_HI, val64));
>> + rp[i].parent_addr_lo =
>> + cpu_to_be32(FIELD_GET(OF_PCI_ADDR_LO, val64));
> cpu_to_be64(res[j].start) would be simpler, but then you'll need
> unaligned accessors.
>
> This all could use some refactoring as you write 3 cell PCI addreses in
> several places. Something like:
>
> of_pci_set_address(__be32 *prop, u64 addr, u32 flags)
Ok. Thanks for pointing this out.
>
>> +
>> + val64 = resource_size(&res[j]);
>> + rp[i].size_hi = cpu_to_be32(FIELD_GET(OF_PCI_SIZE_HI, val64));
>> + rp[i].size_lo = cpu_to_be32(FIELD_GET(OF_PCI_SIZE_LO, val64));
>> +
>> + rp[i].child_addr_hi = rp[i].parent_addr_hi;
>> + rp[i].child_addr_mid = rp[i].parent_addr_mid;
>> + rp[i].child_addr_lo = rp[i].parent_addr_lo;
>> + i++;
>> + }
>> +
>> + ret = of_changeset_add_prop_u32_array(ocs, np, "ranges", (u32 *)rp,
>> + i * sizeof(*rp) / sizeof(u32));
>> +
>> + return ret;
>> +}
>> +
>> +static int of_pci_prop_reg(struct pci_dev *pdev, struct of_changeset *ocs,
>> + struct device_node *np)
>> +{
>> + struct of_pci_addr_pair *reg;
>> + int i = 1, resno, ret = 0;
>> + u32 reg_val, base_addr;
>> + resource_size_t sz;
>> +
>> + reg = kzalloc(sizeof(*reg) * (PCI_STD_NUM_BARS + 1), GFP_KERNEL);
>> + if (!reg)
>> + return -ENOMEM;
>> +
>> + reg_val = FIELD_PREP(OF_PCI_ADDR_FIELD_SS, OF_PCI_ADDR_SPACE_CONFIG) |
>> + FIELD_PREP(OF_PCI_ADDR_FIELD_BUS, pdev->bus->number) |
>> + FIELD_PREP(OF_PCI_ADDR_FIELD_DEV, PCI_SLOT(pdev->devfn)) |
>> + FIELD_PREP(OF_PCI_ADDR_FIELD_FUNC, PCI_FUNC(pdev->devfn));
>> + reg[0].phys_hi = cpu_to_be32(reg_val);
>> +
>> + base_addr = PCI_BASE_ADDRESS_0;
>> + for (resno = PCI_STD_RESOURCES; resno <= PCI_STD_RESOURCE_END;
>> + resno++, base_addr += 4) {
>> + sz = pci_resource_len(pdev, resno);
>> + if (!sz)
>> + continue;
>> +
>> + ret = of_pci_set_addr_flags(&pdev->resource[resno], ®_val);
>> + if (!ret)
>> + continue;
>> +
>> + reg_val &= ~OF_PCI_ADDR_FIELD_REG;
>> + reg_val |= FIELD_PREP(OF_PCI_ADDR_FIELD_REG, base_addr);
>> + reg[i].phys_hi = cpu_to_be32(reg_val);
>> + reg[i].size_hi = cpu_to_be32(FIELD_GET(OF_PCI_SIZE_HI, sz));
>> + reg[i].size_lo = cpu_to_be32(FIELD_GET(OF_PCI_SIZE_LO, sz));
>> + i++;
>> + }
>> +
>> + ret = of_changeset_add_prop_u32_array(ocs, np, "reg", (u32 *)reg,
>> + i * sizeof(*reg) / sizeof(u32));
>> + kfree(reg);
>> +
>> + return ret;
>> +}
>> +
>> +static int of_pci_prop_compatible(struct pci_dev *pdev,
>> + struct of_changeset *ocs,
>> + struct device_node *np)
>> +{
>> + const char *compat_strs[PROP_COMPAT_NUM] = { 0 };
>> + int i, ret;
>> +
>> + compat_strs[PROP_COMPAT_PCI_VVVV_DDDD] =
>> + kasprintf(GFP_KERNEL, "pci%x,%x", pdev->vendor, pdev->device);
>> + compat_strs[PROP_COMPAT_PCICLASS_CCSSPP] =
>> + kasprintf(GFP_KERNEL, "pciclass,%06x", pdev->class);
>> + compat_strs[PROP_COMPAT_PCICLASS_CCSS] =
>> + kasprintf(GFP_KERNEL, "pciclass,%04x", pdev->class >> 8);
>> +
>> + ret = of_changeset_add_prop_string_array(ocs, np, "compatible",
>> + compat_strs, PROP_COMPAT_NUM);
>> + for (i = 0; i < PROP_COMPAT_NUM; i++)
>> + kfree(compat_strs[i]);
>> +
>> + return ret;
>> +}
>> +
>> +static int (*of_pci_endpoint_props[])(struct pci_dev *pdev,
>> + struct of_changeset *ocs,
>> + struct device_node *np) = {
>> + of_pci_prop_reg,
>> + of_pci_prop_compatible,
>> + NULL
>> +};
>> +
>> +static int (*of_pci_bridge_props[])(struct pci_dev *pdev,
>> + struct of_changeset *ocs,
>> + struct device_node *np) = {
>> + of_pci_prop_device_type,
>> + of_pci_prop_address_cells,
>> + of_pci_prop_size_cells,
>> + of_pci_prop_ranges,
>> + of_pci_prop_reg,
>> + of_pci_prop_compatible,
>> + NULL
>> +};
>> +
>> +int of_pci_add_properties(struct pci_dev *pdev, struct of_changeset *ocs,
>> + struct device_node *np)
>> +{
>> + int (**prop_func)(struct pci_dev *pdev, struct of_changeset *ocs,
>> + struct device_node *np);
>> + int i, ret;
>> +
>> + if (pci_is_bridge(pdev))
>> + prop_func = of_pci_bridge_props;
> Skip all this indirection and just make everything direct function
> calls.
ok
Thanks,
Lizhi
>
>> + else
>> + prop_func = of_pci_endpoint_props;
>> +
>> + for (i = 0; prop_func[i]; i++) {
>> + ret = prop_func[i](pdev, ocs, np);
>> + if (ret) {
>> + /*
>> + * The added properties will be released when the
>> + * changeset is destroyed.
>> + */
>> + return ret;
>> + }
>> + }
>> +
>> + return 0;
>> +}
>> diff --git a/drivers/pci/pci-driver.c b/drivers/pci/pci-driver.c
>> index 49238ddd39ee..1540c4c9a770 100644
>> --- a/drivers/pci/pci-driver.c
>> +++ b/drivers/pci/pci-driver.c
>> @@ -1628,7 +1628,8 @@ static int pci_dma_configure(struct device *dev)
>> bridge = pci_get_host_bridge_device(to_pci_dev(dev));
>>
>> if (IS_ENABLED(CONFIG_OF) && bridge->parent &&
>> - bridge->parent->of_node) {
>> + bridge->parent->of_node &&
>> + !of_node_check_flag(bridge->parent->of_node, OF_DYNAMIC)) {
>> ret = of_dma_configure(dev, bridge->parent->of_node, true);
>> } else if (has_acpi_companion(bridge)) {
>> struct acpi_device *adev = to_acpi_device_node(bridge->fwnode);
>> diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
>> index 785f31086313..bd81dc4ca04f 100644
>> --- a/drivers/pci/pci.h
>> +++ b/drivers/pci/pci.h
>> @@ -678,6 +678,25 @@ static inline int devm_of_pci_bridge_init(struct device *dev, struct pci_host_br
>>
>> #endif /* CONFIG_OF */
>>
>> +struct of_changeset;
>> +
>> +#ifdef CONFIG_PCI_DYNAMIC_OF_NODES
>> +void of_pci_make_dev_node(struct pci_dev *pdev);
>> +void of_pci_remove_node(struct pci_dev *pdev);
>> +int of_pci_add_properties(struct pci_dev *pdev, struct of_changeset *ocs,
>> + struct device_node *np);
>> +#else
>> +static inline void
>> +of_pci_make_dev_node(struct pci_dev *pdev)
>> +{
>> +}
>> +
>> +static inline void
>> +of_pci_remove_node(struct pci_dev *pdev)
>> +{
>> +}
>> +#endif /* CONFIG_PCI_DYNAMIC_OF_NODES */
>> +
>> #ifdef CONFIG_PCIEAER
>> void pci_no_aer(void);
>> void pci_aer_init(struct pci_dev *dev);
>> diff --git a/drivers/pci/remove.c b/drivers/pci/remove.c
>> index 4c54c75050dc..0eaa9d9a3609 100644
>> --- a/drivers/pci/remove.c
>> +++ b/drivers/pci/remove.c
>> @@ -23,6 +23,7 @@ static void pci_stop_dev(struct pci_dev *dev)
>> device_release_driver(&dev->dev);
>> pci_proc_detach_device(dev);
>> pci_remove_sysfs_dev_files(dev);
>> + of_pci_remove_node(dev);
>>
>> pci_dev_assign_added(dev, false);
>> }
>> --
>> 2.17.1
>>
>>
On Mon, Dec 5, 2022 at 6:30 PM Lizhi Hou <[email protected]> wrote:
>
>
> On 12/1/22 13:12, Rob Herring wrote:
> > On Mon, Nov 21, 2022 at 08:43:03AM -0800, Lizhi Hou wrote:
> >> The PCI endpoint device such as Xilinx Alveo PCI card maps the register
> >> spaces from multiple hardware peripherals to its PCI BAR. Normally,
> >> the PCI core discovers devices and BARs using the PCI enumeration process.
> >> There is no infrastructure to discover the hardware peripherals that are
> >> present in a PCI device, and which can be accessed through the PCI BARs.
> >>
> >> For Alveo PCI card, the card firmware provides a flattened device tree to
> >> describe the hardware peripherals on its BARs. The Alveo card driver can
> >> load this flattened device tree and leverage device tree framework to
> >> generate platform devices for the hardware peripherals eventually.
> >>
> >> Apparently, the device tree framework requires a device tree node for the
> >> PCI device. Thus, it can generate the device tree nodes for hardware
> >> peripherals underneath. Because PCI is self discoverable bus, there might
> >> not be a device tree node created for PCI devices. This patch is to add
> >> support to generate device tree node for PCI devices.
> >>
> >> Added a kernel option. When the option is turned on, the kernel will
> >> generate device tree nodes for PCI bridges unconditionally.
> >>
> >> Initially, the basic properties are added for the dynamically generated
> >> device tree nodes.
> >>
> >> Signed-off-by: Lizhi Hou <[email protected]>
> >> Signed-off-by: Sonal Santan <[email protected]>
> >> Signed-off-by: Max Zhen <[email protected]>
> >> Reviewed-by: Brian Xu <[email protected]>
> >> ---
> >> drivers/pci/Kconfig | 12 ++
> >> drivers/pci/Makefile | 1 +
> >> drivers/pci/bus.c | 2 +
> >> drivers/pci/msi/irqdomain.c | 6 +-
> >> drivers/pci/of.c | 71 ++++++++++
> >> drivers/pci/of_property.c | 256 ++++++++++++++++++++++++++++++++++++
> >> drivers/pci/pci-driver.c | 3 +-
> >> drivers/pci/pci.h | 19 +++
> >> drivers/pci/remove.c | 1 +
> >> 9 files changed, 368 insertions(+), 3 deletions(-)
> >> create mode 100644 drivers/pci/of_property.c
> >>
> >> diff --git a/drivers/pci/Kconfig b/drivers/pci/Kconfig
> >> index 55c028af4bd9..126c31b79718 100644
> >> --- a/drivers/pci/Kconfig
> >> +++ b/drivers/pci/Kconfig
> >> @@ -198,6 +198,18 @@ config PCI_HYPERV
> >> The PCI device frontend driver allows the kernel to import arbitrary
> >> PCI devices from a PCI backend to support PCI driver domains.
> >>
> >> +config PCI_DYNAMIC_OF_NODES
> >> + bool "Device tree node for PCI devices"
> > Create Devicetree nodes for PCI devices
> Sure.
> >
> > But as I've said before, making this a config option doesn't really work
> > except as something to experiment with. Once you add your driver and
> > want to do a 'select PCI_DYNAMIC_OF_NODES', you've affected everyone
> > else.
>
> Do you mean we should remove PCI_DYNAMIC_OF_NODES and make
> creating dynamic tree node default?
No, I'm saying as long as it is a config option, it's not useful for
more than experimentation. A distro kernel has to decide how to set a
config option for *everyone*.
> Based on the previous discussions, the approach I am implementing
> is to create device tree nodes for all PCI bridges and devices defined
> by pci quirks. I believe a PCI endpoint which is not defined by PCI quirks
> should not to be affected because there is no device tree node is created
> for it.
>
> Are you fine with this approach?
How does that work? The quirks run when a device is discovered. At
that time you've already discovered and probed everything upstream of
the device. So the only thing controlling the upstream devices getting
a DT node is the config option, right?
> >> + depends on OF
> >> + select OF_DYNAMIC
> >> + help
> >> + This option enables support for generating device tree nodes for some
> >> + PCI devices. Thus, the driver of this kind can load and overlay
> >> + flattened device tree for its downstream devices.
> >> +
> >> + Once this option is selected, the device tree nodes will be generated
> >> + for all PCI bridges.
> >> +
> >> choice
> >> prompt "PCI Express hierarchy optimization setting"
> >> default PCIE_BUS_DEFAULT
> >> diff --git a/drivers/pci/Makefile b/drivers/pci/Makefile
> >> index 2680e4c92f0a..cc8b4e01e29d 100644
> >> --- a/drivers/pci/Makefile
> >> +++ b/drivers/pci/Makefile
> >> @@ -32,6 +32,7 @@ obj-$(CONFIG_PCI_P2PDMA) += p2pdma.o
> >> obj-$(CONFIG_XEN_PCIDEV_FRONTEND) += xen-pcifront.o
> >> obj-$(CONFIG_VGA_ARB) += vgaarb.o
> >> obj-$(CONFIG_PCI_DOE) += doe.o
> >> +obj-$(CONFIG_PCI_DYNAMIC_OF_NODES) += of_property.o
> >>
> >> # Endpoint library must be initialized before its users
> >> obj-$(CONFIG_PCI_ENDPOINT) += endpoint/
> >> diff --git a/drivers/pci/bus.c b/drivers/pci/bus.c
> >> index 3cef835b375f..8507cc32b61d 100644
> >> --- a/drivers/pci/bus.c
> >> +++ b/drivers/pci/bus.c
> >> @@ -316,6 +316,8 @@ void pci_bus_add_device(struct pci_dev *dev)
> >> */
> >> pcibios_bus_add_device(dev);
> >> pci_fixup_device(pci_fixup_final, dev);
> >> + if (pci_is_bridge(dev))
> >> + of_pci_make_dev_node(dev);
> >> pci_create_sysfs_dev_files(dev);
> >> pci_proc_attach_device(dev);
> >> pci_bridge_d3_update(dev);
> >> diff --git a/drivers/pci/msi/irqdomain.c b/drivers/pci/msi/irqdomain.c
> >> index e9cf318e6670..eeaf44169bfd 100644
> >> --- a/drivers/pci/msi/irqdomain.c
> >> +++ b/drivers/pci/msi/irqdomain.c
> >> @@ -230,8 +230,10 @@ u32 pci_msi_domain_get_msi_rid(struct irq_domain *domain, struct pci_dev *pdev)
> >> pci_for_each_dma_alias(pdev, get_msi_id_cb, &rid);
> >>
> >> of_node = irq_domain_get_of_node(domain);
> >> - rid = of_node ? of_msi_map_id(&pdev->dev, of_node, rid) :
> >> - iort_msi_map_id(&pdev->dev, rid);
> >> + if (of_node && !of_node_check_flag(of_node, OF_DYNAMIC))
> >> + rid = of_msi_map_id(&pdev->dev, of_node, rid);
> >> + else
> >> + rid = iort_msi_map_id(&pdev->dev, rid);
> > I have no idea if this works. It looks kind of broken already if
> > !of_node calls iort_msi_map_id(). With a DT only system, I would think
> > we'd always call of_msi_map_id(). Have you tested MSIs?
> >
> > With a mixed system, I have no idea what happens. That needs to be
> > understood for MSI, DMA, and interrupts.
>
> Yes, I tested MSI in VM.
>
> # cat
> /sys/devices/platform/3f000000.pcie/pci0000:00/0000:00:02.0/0000:09:00.0/0000:0a:00.0/msi_irqs/29
>
> msi
> # cat /proc/interrupts | grep 29
> 29: 1 0 MSI 5242880 Edge pciehp
>
> The idea is to preserve the current behaviror.
>
> current: PCI device does not have dt node, thus iort_msi_map_id()
> is called.
>
> modified code: PCI device has dt node but with OF_DYNAMIC flag,
> thus iort_msi_map_id() is called.
>
> I was planning to take on of_msi_map_id() for dynamically generated dt node
>
> in future when we see a real use case?
>
> >
> >>
> >> return rid;
> >> }
> >> diff --git a/drivers/pci/of.c b/drivers/pci/of.c
> >> index 196834ed44fe..fb60b04f0b93 100644
> >> --- a/drivers/pci/of.c
> >> +++ b/drivers/pci/of.c
> >> @@ -469,6 +469,8 @@ static int of_irq_parse_pci(const struct pci_dev *pdev, struct of_phandle_args *
> >> } else {
> >> /* We found a P2P bridge, check if it has a node */
> >> ppnode = pci_device_to_OF_node(ppdev);
> >> + if (of_node_check_flag(ppnode, OF_DYNAMIC))
> >> + ppnode = NULL;
> >> }
> >>
> >> /*
> >> @@ -599,6 +601,75 @@ int devm_of_pci_bridge_init(struct device *dev, struct pci_host_bridge *bridge)
> >> return pci_parse_request_of_pci_ranges(dev, bridge);
> >> }
> >>
> >> +#if IS_ENABLED(CONFIG_PCI_DYNAMIC_OF_NODES)
> >> +
> >> +void of_pci_remove_node(struct pci_dev *pdev)
> >> +{
> >> + struct device_node *dt_node;
> > node or np are the typical names.
> Will fix this.
> >
> >> +
> >> + dt_node = pci_device_to_OF_node(pdev);
> >> + if (!dt_node || !of_node_check_flag(dt_node, OF_DYNAMIC))
> >> + return;
> >> + pdev->dev.of_node = NULL;
> >> +
> >> + of_destroy_node(dt_node);
> >> +}
> >> +
> >> +void of_pci_make_dev_node(struct pci_dev *pdev)
> >> +{
> >> + struct device_node *parent, *dt_node = NULL;
> >> + const char *pci_type = "dev";
> >> + struct of_changeset *cset;
> >> + const char *full_name;
> >> + int ret;
> >> +
> >> + /*
> >> + * If there is already a device tree node linked to this device,
> >> + * return immediately.
> >> + */
> >> + if (pci_device_to_OF_node(pdev))
> >> + return;
> >> +
> >> + /* Check if there is device tree node for parent device */
> >> + if (!pdev->bus->self)
> >> + parent = pdev->bus->dev.of_node;
> >> + else
> >> + parent = pdev->bus->self->dev.of_node;
> >> + if (!parent)
> >> + return;
> >> +
> >> + if (pci_is_bridge(pdev))
> >> + pci_type = "pci";
> > What's the node name if not a bridge? I don't see how that would work.
> >
> > It should depend on the device class if it has one.
>
> pci_type is initialized with "dev"
>
> + const char *pci_type = "dev";
I missed that...
> Do you mean I should use class instead of pci_is_bridge()?
>
> if ((pdev->class >> 24) == PCI_BASE_CLASS_BRIDGE)
> pci_type = "pci";
Well, maybe as preparation to support other classes. If you had a UART
for example, the node name is 'serial'. I don't think you need to
worry about those ATM. We may need some way for the name to come from
the driver as not all devices have a class. Yours for example, we'd
want something like 'fpga@...' ideally. Maybe that's fine to leave as
a known issue.
Rob
On 12/7/22 14:44, Rob Herring wrote:
> On Mon, Dec 5, 2022 at 6:30 PM Lizhi Hou <[email protected]> wrote:
>>
>> On 12/1/22 13:12, Rob Herring wrote:
>>> On Mon, Nov 21, 2022 at 08:43:03AM -0800, Lizhi Hou wrote:
>>>> The PCI endpoint device such as Xilinx Alveo PCI card maps the register
>>>> spaces from multiple hardware peripherals to its PCI BAR. Normally,
>>>> the PCI core discovers devices and BARs using the PCI enumeration process.
>>>> There is no infrastructure to discover the hardware peripherals that are
>>>> present in a PCI device, and which can be accessed through the PCI BARs.
>>>>
>>>> For Alveo PCI card, the card firmware provides a flattened device tree to
>>>> describe the hardware peripherals on its BARs. The Alveo card driver can
>>>> load this flattened device tree and leverage device tree framework to
>>>> generate platform devices for the hardware peripherals eventually.
>>>>
>>>> Apparently, the device tree framework requires a device tree node for the
>>>> PCI device. Thus, it can generate the device tree nodes for hardware
>>>> peripherals underneath. Because PCI is self discoverable bus, there might
>>>> not be a device tree node created for PCI devices. This patch is to add
>>>> support to generate device tree node for PCI devices.
>>>>
>>>> Added a kernel option. When the option is turned on, the kernel will
>>>> generate device tree nodes for PCI bridges unconditionally.
>>>>
>>>> Initially, the basic properties are added for the dynamically generated
>>>> device tree nodes.
>>>>
>>>> Signed-off-by: Lizhi Hou <[email protected]>
>>>> Signed-off-by: Sonal Santan <[email protected]>
>>>> Signed-off-by: Max Zhen <[email protected]>
>>>> Reviewed-by: Brian Xu <[email protected]>
>>>> ---
>>>> drivers/pci/Kconfig | 12 ++
>>>> drivers/pci/Makefile | 1 +
>>>> drivers/pci/bus.c | 2 +
>>>> drivers/pci/msi/irqdomain.c | 6 +-
>>>> drivers/pci/of.c | 71 ++++++++++
>>>> drivers/pci/of_property.c | 256 ++++++++++++++++++++++++++++++++++++
>>>> drivers/pci/pci-driver.c | 3 +-
>>>> drivers/pci/pci.h | 19 +++
>>>> drivers/pci/remove.c | 1 +
>>>> 9 files changed, 368 insertions(+), 3 deletions(-)
>>>> create mode 100644 drivers/pci/of_property.c
>>>>
>>>> diff --git a/drivers/pci/Kconfig b/drivers/pci/Kconfig
>>>> index 55c028af4bd9..126c31b79718 100644
>>>> --- a/drivers/pci/Kconfig
>>>> +++ b/drivers/pci/Kconfig
>>>> @@ -198,6 +198,18 @@ config PCI_HYPERV
>>>> The PCI device frontend driver allows the kernel to import arbitrary
>>>> PCI devices from a PCI backend to support PCI driver domains.
>>>>
>>>> +config PCI_DYNAMIC_OF_NODES
>>>> + bool "Device tree node for PCI devices"
>>> Create Devicetree nodes for PCI devices
>> Sure.
>>> But as I've said before, making this a config option doesn't really work
>>> except as something to experiment with. Once you add your driver and
>>> want to do a 'select PCI_DYNAMIC_OF_NODES', you've affected everyone
>>> else.
>> Do you mean we should remove PCI_DYNAMIC_OF_NODES and make
>> creating dynamic tree node default?
> No, I'm saying as long as it is a config option, it's not useful for
> more than experimentation. A distro kernel has to decide how to set a
> config option for *everyone*.
Ok. I will keep this option.
>
>> Based on the previous discussions, the approach I am implementing
>> is to create device tree nodes for all PCI bridges and devices defined
>> by pci quirks. I believe a PCI endpoint which is not defined by PCI quirks
>> should not to be affected because there is no device tree node is created
>> for it.
>>
>> Are you fine with this approach?
> How does that work? The quirks run when a device is discovered. At
> that time you've already discovered and probed everything upstream of
> the device. So the only thing controlling the upstream devices getting
> a DT node is the config option, right?
Yes, correct.
I meant that the endpoint devices which are not defined in quirks will
not have devicetree node generated. Thus, the behavior of those endpoint
device driver will not be affected.
>
>>>> + depends on OF
>>>> + select OF_DYNAMIC
>>>> + help
>>>> + This option enables support for generating device tree nodes for some
>>>> + PCI devices. Thus, the driver of this kind can load and overlay
>>>> + flattened device tree for its downstream devices.
>>>> +
>>>> + Once this option is selected, the device tree nodes will be generated
>>>> + for all PCI bridges.
>>>> +
>>>> choice
>>>> prompt "PCI Express hierarchy optimization setting"
>>>> default PCIE_BUS_DEFAULT
>>>> diff --git a/drivers/pci/Makefile b/drivers/pci/Makefile
>>>> index 2680e4c92f0a..cc8b4e01e29d 100644
>>>> --- a/drivers/pci/Makefile
>>>> +++ b/drivers/pci/Makefile
>>>> @@ -32,6 +32,7 @@ obj-$(CONFIG_PCI_P2PDMA) += p2pdma.o
>>>> obj-$(CONFIG_XEN_PCIDEV_FRONTEND) += xen-pcifront.o
>>>> obj-$(CONFIG_VGA_ARB) += vgaarb.o
>>>> obj-$(CONFIG_PCI_DOE) += doe.o
>>>> +obj-$(CONFIG_PCI_DYNAMIC_OF_NODES) += of_property.o
>>>>
>>>> # Endpoint library must be initialized before its users
>>>> obj-$(CONFIG_PCI_ENDPOINT) += endpoint/
>>>> diff --git a/drivers/pci/bus.c b/drivers/pci/bus.c
>>>> index 3cef835b375f..8507cc32b61d 100644
>>>> --- a/drivers/pci/bus.c
>>>> +++ b/drivers/pci/bus.c
>>>> @@ -316,6 +316,8 @@ void pci_bus_add_device(struct pci_dev *dev)
>>>> */
>>>> pcibios_bus_add_device(dev);
>>>> pci_fixup_device(pci_fixup_final, dev);
>>>> + if (pci_is_bridge(dev))
>>>> + of_pci_make_dev_node(dev);
>>>> pci_create_sysfs_dev_files(dev);
>>>> pci_proc_attach_device(dev);
>>>> pci_bridge_d3_update(dev);
>>>> diff --git a/drivers/pci/msi/irqdomain.c b/drivers/pci/msi/irqdomain.c
>>>> index e9cf318e6670..eeaf44169bfd 100644
>>>> --- a/drivers/pci/msi/irqdomain.c
>>>> +++ b/drivers/pci/msi/irqdomain.c
>>>> @@ -230,8 +230,10 @@ u32 pci_msi_domain_get_msi_rid(struct irq_domain *domain, struct pci_dev *pdev)
>>>> pci_for_each_dma_alias(pdev, get_msi_id_cb, &rid);
>>>>
>>>> of_node = irq_domain_get_of_node(domain);
>>>> - rid = of_node ? of_msi_map_id(&pdev->dev, of_node, rid) :
>>>> - iort_msi_map_id(&pdev->dev, rid);
>>>> + if (of_node && !of_node_check_flag(of_node, OF_DYNAMIC))
>>>> + rid = of_msi_map_id(&pdev->dev, of_node, rid);
>>>> + else
>>>> + rid = iort_msi_map_id(&pdev->dev, rid);
>>> I have no idea if this works. It looks kind of broken already if
>>> !of_node calls iort_msi_map_id(). With a DT only system, I would think
>>> we'd always call of_msi_map_id(). Have you tested MSIs?
>>>
>>> With a mixed system, I have no idea what happens. That needs to be
>>> understood for MSI, DMA, and interrupts.
>> Yes, I tested MSI in VM.
>>
>> # cat
>> /sys/devices/platform/3f000000.pcie/pci0000:00/0000:00:02.0/0000:09:00.0/0000:0a:00.0/msi_irqs/29
>>
>> msi
>> # cat /proc/interrupts | grep 29
>> 29: 1 0 MSI 5242880 Edge pciehp
>>
>> The idea is to preserve the current behaviror.
>>
>> current: PCI device does not have dt node, thus iort_msi_map_id()
>> is called.
>>
>> modified code: PCI device has dt node but with OF_DYNAMIC flag,
>> thus iort_msi_map_id() is called.
>>
>> I was planning to take on of_msi_map_id() for dynamically generated dt node
>>
>> in future when we see a real use case?
>>
>>>> return rid;
>>>> }
>>>> diff --git a/drivers/pci/of.c b/drivers/pci/of.c
>>>> index 196834ed44fe..fb60b04f0b93 100644
>>>> --- a/drivers/pci/of.c
>>>> +++ b/drivers/pci/of.c
>>>> @@ -469,6 +469,8 @@ static int of_irq_parse_pci(const struct pci_dev *pdev, struct of_phandle_args *
>>>> } else {
>>>> /* We found a P2P bridge, check if it has a node */
>>>> ppnode = pci_device_to_OF_node(ppdev);
>>>> + if (of_node_check_flag(ppnode, OF_DYNAMIC))
>>>> + ppnode = NULL;
>>>> }
>>>>
>>>> /*
>>>> @@ -599,6 +601,75 @@ int devm_of_pci_bridge_init(struct device *dev, struct pci_host_bridge *bridge)
>>>> return pci_parse_request_of_pci_ranges(dev, bridge);
>>>> }
>>>>
>>>> +#if IS_ENABLED(CONFIG_PCI_DYNAMIC_OF_NODES)
>>>> +
>>>> +void of_pci_remove_node(struct pci_dev *pdev)
>>>> +{
>>>> + struct device_node *dt_node;
>>> node or np are the typical names.
>> Will fix this.
>>>> +
>>>> + dt_node = pci_device_to_OF_node(pdev);
>>>> + if (!dt_node || !of_node_check_flag(dt_node, OF_DYNAMIC))
>>>> + return;
>>>> + pdev->dev.of_node = NULL;
>>>> +
>>>> + of_destroy_node(dt_node);
>>>> +}
>>>> +
>>>> +void of_pci_make_dev_node(struct pci_dev *pdev)
>>>> +{
>>>> + struct device_node *parent, *dt_node = NULL;
>>>> + const char *pci_type = "dev";
>>>> + struct of_changeset *cset;
>>>> + const char *full_name;
>>>> + int ret;
>>>> +
>>>> + /*
>>>> + * If there is already a device tree node linked to this device,
>>>> + * return immediately.
>>>> + */
>>>> + if (pci_device_to_OF_node(pdev))
>>>> + return;
>>>> +
>>>> + /* Check if there is device tree node for parent device */
>>>> + if (!pdev->bus->self)
>>>> + parent = pdev->bus->dev.of_node;
>>>> + else
>>>> + parent = pdev->bus->self->dev.of_node;
>>>> + if (!parent)
>>>> + return;
>>>> +
>>>> + if (pci_is_bridge(pdev))
>>>> + pci_type = "pci";
>>> What's the node name if not a bridge? I don't see how that would work.
>>>
>>> It should depend on the device class if it has one.
>> pci_type is initialized with "dev"
>>
>> + const char *pci_type = "dev";
> I missed that...
>
>> Do you mean I should use class instead of pci_is_bridge()?
>>
>> if ((pdev->class >> 24) == PCI_BASE_CLASS_BRIDGE)
>> pci_type = "pci";
> Well, maybe as preparation to support other classes. If you had a UART
> for example, the node name is 'serial'. I don't think you need to
> worry about those ATM. We may need some way for the name to come from
> the driver as not all devices have a class. Yours for example, we'd
> want something like 'fpga@...' ideally. Maybe that's fine to leave as
> a known issue.
Got it. I will keep generic endpoint name 'dev' for now. Thanks.
Lizhi
>
> Rob